In general, multi-party conference calls enable parties at different locations to communicate with each other. In the current paradigm, in order to support a conference call for n parties, n(n−1) media streams must be established. Unfortunately, for n>2, the bandwidth consumed by the n(n−1) media streams is significantly larger then necessary for supporting communications between the n parties. Similarly, in order to support a conference call for n parties using a standard conference bridge, 2n media streams must be established. Unfortunately, when a portion of the parties are located in a sub-network, and the conference bridge is located outside that sub-network, the network resources consumed by the 2n media streams is significantly larger then necessary for supporting communications between the n parties.